# Conflicts: # docs/notes/server-stats.md
1.1 KiB
Server stats
Send a server_stats websocket request to receive a server_stats_result with
host, CPU, memory, storage, Pipelines data, process, and best-effort GPU health.
Unsupported or unavailable sections include an error or reason rather than
failing the complete response.
Client
Settings → Server health shows the machine's CPU, memory and swap, GPU, local mounts, Pipelines data directory and server process. Native Material and Cupertino controls follow the active host. Meters use the theme primary colour, high usage from 80%, and critical from 95%. Missing measurements display “—”; section and mount errors remain visible alongside available sections.
Auto-refresh samples every five seconds while the route is current and the app is active, stops when covered or disposed, and never overlaps a pending request. Manual refresh and an auto-refresh toggle are available in the host summary.
The full-app capture fixture answers server_stats from a shared snapshot with
one GPU and two mounts. Open /settings and choose Server health; fixture
captures do not prove live host metrics or device behaviour.